Use of cell doublets for studying cytokinesis regulation reveals a new form of cytokinesis regression


 Cytokinesis mediates separation of daughter cells at the end of cell division. We have developed a high-throughput approach for monitoring cell-autonomous cytokinesis in non-adherent cells. Focusing on cytokinesis termination, we show that chemical inhibition of protein phosphatase 1 (PP1) and PP2A specifically in late cytokinesis activates cytokinesis regression, which is distinct from any known cytokinesis failure, and is not a by-product of abnormal furrow ingression or chromatin bridges. This process is characterized by the formation of cortical blebs primarily at the intercellular bridge, reopening of the cleavage furrow and reassembly of an interphase-like microtubule network, but not by chromatin recondensation and mitotic spindle formation. Finally, cytokinesis regression is suppressed by chemical inhibition of aurora kinases but not Cdk1 or PLK1. Altogether, our results highlight a hitherto uncharacterized facet of the counter-activity of PP1/PP2A and aurora kinases in the final step of cell division, which ultimately secure the conclusion of cytokinesis, thereby preventing polyploidy and genomic instability.


Introduction 43
Scientists have been fascinated by the unidirectionality of the cell cycle ever since 44 the pioneering era of cell biology 1 . Cell cycle unidirectionality is largely dependent on 45 orderly phosphorylation, dephosphorylation and ubiquitin-mediated proteolysis of 46 regulatory and structural proteins 2 . Mitotic entry is driven by the kinase activity of CDK1-47 cyclin B 3-7 . Following mitotic entry, proteolysis of cyclin B and inactivation of CDK1 8-10 , 48 concomitantly with a massive dephosphorylation wave executed by protein phosphatases 1 49 (PP1) and PP2A and by more canonical cell cycle phosphatases (e.g., CDC14) orchestrate cell 50 division and mitotic exit into the G1 phase (G1) of the following cell cycle [11][12][13][14][15][16][17][18][19] . This process is 51 essential for successful cell proliferation. 52 Cell division ends with cytokinesis, in which the content of a mother cell is 53 partitioned between two daughter cells. The physical division of the mother cell after 54 chromosome segregation is first and foremost mediated by the contractile actomyosin ring, 55 which ingresses a cleavage furrow at the cell equator. Furrow ingression ends with the 56 formation of a dense filamentous structure known as the midbody 20,21 . Cytokinesis is 57 concluded when the intercellular bridge that connects the two sister cells is cut. This 58 process is known as 'cytokinetic abscission' (abscission), is cell-cycle coordinated 22-24 and 59 integrates external cues from cell-cell and cell-surface contacts. These cues may, for 60 example, involve forces that shape the dividing cell and the intercellular bridge in ways that 61 influence the timing and mode of cytokinesis termination [25][26][27] .

Inhibition of PP1/PP2A in late cytokinesis induces cytokinesis regression 131
The phosphatases PP1 and PP2A are essential for the dephosphorylation wave that 132 regulates mitotic exit and cytokinesis. The activity of PP1/PP2A can be blocked by the small 133 molecule drug calyculin A (CaA) 39,40 . We therefore tested whether CaA affects any of the 134 four cytokinesis indices. Although the phase index was only modestly altered following 1 or 135 2 hrs exposure to CaA, the shift in the other indices was dramatic ( Figures. 1g-h, and S2). 136 The sharp reduction in the doublet index and size index F1 revealed a drop in the fraction of 137 8-shaped cytokinetic cells. Importantly, the noticeable reciprocal increase in size index F2 138 suggested that, rather than dividing, a significant portion of late-cytokinesis cells (F1 index) 139 acquired a more spherical shape during incubation with CaA. Alternatively, the increase in 140 size index F2 might have revealed a more global decoupling between mitotic exit (mAG-141 geminin proteolysis) and cytokinesis (cell division). Overall, the flow cytometry data 142 presented in Figure 1 demonstrate that CaA-induced inhibition of PP1 and/or PP2A impairs 143 cytokinesis. 144 PP1 and PP2A regulate every stage of the cell cycle, including cell division 16 . 145 Chemical inhibition of PP1/PP2A blocks cell division ( Figure S3) 41 . To examine the impact of 146 CaA specifically on cytokinesis termination, we sorted L1210 cells using size index F1 as a 147 gating strategy and obtained late-cytokinesis cells exhibiting a fully ingressed furrow and 148 intercellular bridge (Figures 2a and S4), as well as decondensed chromatin and reassembled 149 nuclei typifying cells at late telophase (Figure 2b). We refer to these cells as 'late-cytokinesis 150 cells' from now on. To clarify, late-cytokinesis cells were isolated without chemical labeling, 151 cell cycle arrest or any other means of pre-synchronization to allow direct and systematic 152 manipulations of cytokinesis termination, thereby avoiding both manipulation of cells pre-153 and during furrow ingression and/or effects of external cues associated with cell-to-cell and 154 cell-to-surface contacts. 155 We treated late-cytokinesis cells with CaA, hypothesizing that inhibition of PP1/PP2A 156 at this stage will affect abscission. Admittedly, we did not foresee the outcome of this 157 manipulation: the thin intercellular bridge puffed up, the furrow regressed completely, and 158 the sister cells coalesced, generating a spherical polyploid cell (Figure 2c and 159 Supplementary movies 1-2). This phenotype looked as if the entire process of cytokinesis 160 proceeded backwards. We therefore named the process 'cytokinesis regression'. Cytokinesis 161 reversal is ubiquitous, dose-dependent and highly reproducible (Figures 2d-e and S5): at 50 162 nM CaA, 80% of the sister cells completed cytokinesis regression (%CR) and coalesced; 163 lowering the dose of CaA to a 5 nM moderately reduced this value to 49%. This lower score 164 was ideal for experiments discovering factors that either facilitate or repress the process. 165 We established the 5 nM CaA treatment as the default protocol and limited image analysis 166 to 3 h, which is when cytokinesis regression reaches a (near) maximum level. 167 Cytokinesis regression is not the result of cell fusion ( Figure S6), nor is it dependent 168 on cell sorting ( Figure S7). Importantly, this phenomenon is not cell-type specific or even 169 limited to non-adherent cells, as late-cytokinesis human epithelial HeLa cells exhibited 170 cytokinesis regression following exposure to 5 nM CaA, with one noticeable difference -171 cell membrane blebbing 42 was considerably more profound in HeLa cells relative to L1210 172 cells or any other lymphoblast cell line we tested (Figures 2f-h Figures 2i and S12). Note that time resolution was greatly limited by 186 phototoxicity, in particular when z-stacks were required for 3D information and/or for 187 detailed deconvolution of spherical objects with ~12-15 µM/diameter. resembles sister nuclei coalescence, which is known in yeast 45 , a closed mitosis organism, 201 but not in mammalian cells. More in-depth live imaging and electron microscopy analyses, 202 which are beyond the scope of this study, will be required to investigate the dynamics of the 203 nucleus during cytokinesis reversal. 204 To summarize, in Figure 2 we show that PP1/PP2A inhibition in late cytokinesis 205 triggers furrow regression followed by cell coalescence. This hitherto unknown process is 206 not an epiphenomenon of long-term manipulations, abnormal cytokinesis or improper 207 furrow ingression induced by chemical drugs or genetically 41,44,46-51 , nor it is induced by 208 chromatin bridges at the intercellular bridge ( Figure S14) and the resulting 'No-Cut' 209 response 38,52 . These results highlight an unappreciated role for PP1/PP2A and overall 210 dephosphorylation in driving cytokinesis to termination. The enhanced blebbing activity 211 associated with cytokinesis reversal likely indicates increased actin-myosin contractility 42 . It 212 is noteworthy that CaA is known to intensify cellular contractility through regulation of 213 myosin II 53,54 , which is the major motor driving furrow ingression 55 . 214 215

Cytokinesis regression is driven by aurora kinases 216
The prevalent understanding of the unidirectionality of mitotic exit is based on the 217 premise that CDK1 activity cannot be restored following degradation of cyclin B, and that 218 cell cycle phosphatases counteract the activity of mitotic kinases to regulate orderly 219 proteolysis and other cascading events. Eventually, this cell cycle 'system' is reset before the 220 cell is becoming committed to cycling again. 221 The cells we selected for the study of cytokinesis termination are at a very late stage 222 of cytokinesis, and thus we did not expect inhibition of already-inactive CDK1 to induce 223 cytokinesis regression by itself or to have any effect on CaA-induced cytokinesis regression. 224 Indeed, incubation of late-cytokinesis cells with the selective CDK1 inhibitor RO-3306 did 225 not induce cytokinesis regression (Figs. 3a and S15), and the effect of CDK1 inhibition on 226 CaA-induced cytokinesis regression was almost negligible, albeit statistically significant (Figs. 227 3b and S16, and supplementary movies 5 and 6). These results suggest that remnant CDK1 228 activity associated with late cytokinesis may still be of some relevance to cytokinesis 229 regression but, overall, when the phosphatase activity of PP1/PP2A is repressed, CDK1 is not 230 the driving force of this process. 231 We hypothesized that other mitotic kinases are necessary for CaA-induced 232 cytokinesis regression and tested the dependency of this process on aurora kinases and on 233 PLK1, which are key regulators of cytokinesis for which reliable small-molecule inhibitors are 234 available. Aurora kinases and PLK1 are degraded during mitotic exit following cyclin B 235 degradation and CDK1 inactivation, and their activity in late cytokinesis is undoubtedly 236 considerable 56 . We isolated late-cytokinesis cells and co-treated them with CaA and one of 237 the following kinase inhibitors: VX-680 (inhibitor aurora A/B/C), AZD-1152 (aurora B) or 238 BI2356 (PLK1). Both VX-680 and AZD-1152 strongly repressed CaA-induced cytokinesis 239 regression ( Fig. 3c-d, 3f and S16, and supplementary movies 7 and 8). The outcome of PLK1 240 inhibition was markedly different: BI2356 not only did not repress cytokinesis regression, it 241 mildly, albeit consistently, enhanced it ( Fig. 3e-f and S16, and supplementary movie 9). 242 Note that neither BI2356 nor VX-680 or AZD-1152 induced cytokinesis regression on their 243 own (Figs. 3a and S15). We concluded that cytokinesis regression induced by PP1/PP2A 244 inhibition is repressed by inactivation of aurora kinases, but not PLK1. These results suggest 245 that by the time cells are about to conclude cytokinesis and undergo abscission, they still 246 possess an inherent ability to withdraw from cell division and block cell propagation. This 247 ability is dependent on the activity of aurora kinases, and is prevented by PP1 and/or PP2A. 248 The counteracting activities of aurora kinases and PP1/PP2A is well established (see for 249 example Refs 57-59 ). We speculate that the phosphatase activity of PP1/PP2A secures the 250 conclusion of cytokinesis by directly repressing the activity of aurora kinases, and potentially 251 of other mitotic kinases, through feedback mechanisms and/or through dephosphorylation 252 of their key substrates during mitotic exit (illustrated in Fig. 3g). 253

254
The phenotype of cytokinesis regression, although regulated by the core machinery 255 of mitotic exit, is fundamentally different from the mitosis reversal reported in the classic 256 study by Gorbsky and Stukenberg 43 . In their study, mitotic exit reversibility was achieved by 257 inhibiting proteasomal degradation of cyclin B or by expressing a non-degradable cyclin B. 258 Mitotic cells advancing through mitotic exit upon drug-induced inactivation of CDK1 were 259 able to re-enter mitosis, because cyclin B1 remained stable. Here, cytokinesis regression is 260 induced considerably after the degradation of endogenous cyclin B has commenced. Indeed, 261 we could not detect mitotic reversal in late-cytokinesis L1210 cells treated with proteasome 262 inhibitors: treatment with 25 µM of MG132 caused 0% cytokinesis regression (N=107), and 263 treatment with 1 µM bortezomib caused <1% cytokinesis regerssion (N=370) at t=3 h. These 264 results suggest that at this advanced stage of cytokinesis, the mitotic activity of cyclin B1-265 CDK1 complex cannot be rescued by inhibition of proteolysis. Thus, we can say with 266 confidence that the mechanisms of mitosis reversal and cytokinesis regression are different. 267 Overall, we demonstrate an inherent decoupling between cytokinesis, i.e., the physical 268 with the aurora kinases inhibitor VX-680 or with DMSO ( Fig. 4b and c). None of the G2/M 285 cells treated with VX-680 divided: about half (54%) of the cells failed to enter cytokinesis, 286 and the cells that did eventually commence cytokinesis (46%), failed to conclude furrow 287 ingression, and re-acquired a spherical shape (Fig. 4b and c, Fig. S17 To summarize, the discovery of CaA-mediated cytokinesis regression is yet another 298 demonstration of the elaborate control of the unidirectionality of the cell cycle, in which 299 PP1/PP2A activity is required for preventing polyploidy and maintaining genome and tissue 300

integrity. 301
Small-scale and large-scale methodologies for studying cytokinesis are almost 302 exclusively based on genetic or chemical manipulation of adherent cells, followed by 303 automated imaging and advanced image analyses. We encourage colleagues who agree 304 with us that mammalian cytokinesis must be studied also in non-adherent cells and in a 305 more cell-autonomous fashion to take full advantage of our new set of flow cytometry tools 306